Bug 1986509

Summary:	libvirt does not pass discard flag to copy-on-read blockdev layers
Product:	Red Hat Enterprise Linux 9	Reporter:	Richard W.M. Jones <rjones>
Component:	libvirt	Assignee:	Peter Krempa <pkrempa>
Status:	CLOSED CURRENTRELEASE	QA Contact:	Han Han <hhan>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	9.0	CC:	jdenemar, lmen, mxie, pkrempa, tzheng, virt-maint, xuzhang
Target Milestone:	beta	Keywords:	Triaged
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:	libvirt-7.6.0-1.el9	Doc Type:	If docs needed, set a value
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2021-12-07 21:57:54 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:	7.6.0
Embargoed:
Bug Depends On:
Bug Blocks:	910269

Description Richard W.M. Jones 2021-07-27 17:42:23 UTC

Description of problem:

The long description is here:
https://listman.redhat.com/archives/libguestfs/2021-July/msg00067.html

The short description is: When we use the copy_on_read="on"
and discard="unmap" attributes together on a <disk/> element,
libvirt does not set up blockdev correctly, so the unmap
flag is dropped.  This results in trim messages being
dropped in the copy-on-read layer.

Version-Release number of selected component (if applicable):

libvirt-7.5.0-1.el9

How reproducible:

100%

Steps to Reproduce:

See https://listman.redhat.com/archives/libguestfs/2021-July/msg00067.html
for a full reproducer.

Peter has suggested this patch to libvirt:
https://gitlab.com/pipo.sk/libvirt/-/commit/13d30ec012b258105678356e504c0f5e79881956

I tested Peter's patch and it works for me.

Comment 1 Richard W.M. Jones 2021-07-28 09:41:19 UTC

Upstream fix:
https://listman.redhat.com/archives/libvir-list/2021-July/thread.html#00823

Comment 2 Peter Krempa 2021-07-28 13:22:05 UTC

Fixed upstream:

commit dc0b9c8376afb88eb3a3156d7e991d29d9c196a8
Author: Peter Krempa <pkrempa>
Date:   Tue Jul 27 19:12:09 2021 +0200

    qemu: block: Pass discard requests through the copy-on-read block filter
    
    We need to pass the 'trim' requests through the copy-on-read filter so
    if a user configures a discard policy on the disk the requests get
    through to the appropriate format layer in the blockdev tree.
    
    Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=1986509
    Reported-by: Richard W.M. Jones <rjones>
    Tested-by: Richard W.M. Jones <rjones>
    Signed-off-by: Peter Krempa <pkrempa>
    Reviewed-by: Daniel P. Berrangé <berrange>
    Reviewed-by: Michal Privoznik <mprivozn>
    Reviewed-by: Richard W.M. Jones <rjones>

v7.6.0-rc1-4-gdc0b9c8376

Comment 3 Han Han 2021-08-02 02:24:41 UTC

Reproduced on libvirt-7.5.0-1.module+el8.5.0+11664+59f87560.x86_64:
For an VM with XML:
<domain>
...
  <disk>
  ...
      <driver name='qemu' type='qcow2' copy_on_read='on' discard='unmap'/>
  ...
  </disk>
...
</domain>


Extract its qemu cmdline by `virsh domxml-to-native --domain VM qemu-argv`
The qemu cmdline of -blockdev is:
-blockdev '{"driver":"gluster","volume":"gv0","path":"hhan.qcow2","server":[{"type":"inet","host":"10.0.148.14","port":"24007"}],"debug":4,"node-name":"libvirt-1-storage","auto-read-only":true,"discard":"unmap"}' -blockdev '{"node-name":"libvirt-1-format","read-only":false,"discard":"unmap","driver":"qcow2","file":"libvirt-1-storage"}' -blockdev '{"driver":"copy-on-read","node-name":"libvirt-CoR-vda","file":"libvirt-1-format"}'

The "discard":"unmap" is not in the copy-on-read layer.


The libvirt upstream has covered this unit test.

Comment 6 Han Han 2021-08-23 02:11:50 UTC

Tested on libvirt-7.6.0-2.el9.x86_64 as comment3:
Results:
-blockdev {"node-name":"libvirt-1-format","read-only":false,"discard":"unmap","driver":"qcow2","file":"libvirt-1-storage","backing":null} -blockdev {"driver":"copy-on-read","node-name":"libvirt-CoR-sda","file":"libvirt-1-format","discard":"unmap"}

"discard":"unmap" is in copy-on-read blockdev node.